README.md

metadata

language:
  - en
license: mit
pretty_name: Cancer Abstract Dataset
size_categories:
  - 1K\<n\<10K
tags:
  - biomedical
  - oncology
  - cancer
  - text-classification
  - nlp
  - graph-neural-networks
  - document-classification
task_categories:
  - text-classification

Cancer Abstract Dataset

Dataset Summary

The Cancer Abstract Dataset is a curated collection of biomedical research abstracts categorized by cancer type. It was developed to support research in document classification, low-resource biomedical NLP, and graph-based deep learning approaches.

This dataset was introduced in the following publication:

Hossain, E., Nuzhat, T., Masum, S., et al.
**R-GAT: cancer document classification leveraging graph-based residual network for scenarios with limited data.**
Scientific Reports, 16, 6582 (2026).
https://doi.org/10.1038/s41598-026-39894-6

---

Dataset Description

This dataset contains categorized research abstracts related to major cancer types. It is suitable for:

• Biomedical text classification
• Topic modeling
• Low-resource learning experiments
• Graph-based NLP methods
• Transformer-based fine-tuning
• Benchmarking uncertainty-aware LLMs

---

Dataset Structure

Total Samples

1,874 abstracts

Format

CSV (Comma-Separated Values)

Fields

Field Description

---

Abstract Full research abstract text Category Cancer type label

Categories

• Lung_Cancer
• Thyroid_Cancer
• Colon_Cancer
• Generic

---

Example Usage

from datasets import load_dataset

dataset = load_dataset("EliasHossain/CancerAbstracts")

print(dataset["train"][0])

---

Intended Use

The dataset is intended for:

• Supervised text classification
• Graph neural network research
• Transformer-based fine-tuning
• Biomedical NLP benchmarking
• Limited-data learning evaluation

This dataset is not intended for clinical decision-making.

---

Data Collection and Processing

Abstracts were curated and categorized for research purposes in oncology-related document classification experiments. Standard preprocessing steps were applied to ensure formatting consistency.

No personally identifiable information (PII) or protected health information (PHI) is included.

---

Citation

If you use this dataset, please cite:

@article{hossain2026rgat,
  title={R-GAT: cancer document classification leveraging graph-based residual network for scenarios with limited data},
  author={Hossain, Elias and Nuzhat, Tasfia and Masum, S. and others},
  journal={Scientific Reports},
  volume={16},
  pages={6582},
  year={2026},
  doi={10.1038/s41598-026-39894-6}
}

---

Contributors

• Elias Hossain
  Mississippi State University, USA

• Tasfia Nuzhat
  Chittagong Independent University, Bangladesh

---

License

MIT License